Long-simmering resource nationalism in Tanzania heated up this March with a sudden export ban on ore concentrates by Tanzanian President John Magufuli, who said that “based on the information that I have, if I say what is really inside these containers, it could make any patriotic Tanzanian cry … from now onwards, no mineralized sand will be exported from Tanzania … there is no country being robbed of its mineral wealth like Tanzania.”
It then erupted in June with the release of reports by two presidentially appointed committees alleging that gold miner and Barrick Gold subsidiary Acacia Mining was defrauding the Tanzanian government of tax revenue and royalties by substantially understating the value of its exported concentrates.
The committee reports confirmed the president’s position and were enthusiastically amplified by Tanzanian Swahili newspapers, but solidly refuted by Acacia and even international observers sympathetic to the cause of resource nationalism.
However, one Tanzanian weekly, Mawio, stepped out of line by examining the role of past presidents in resource deals and was promptly shut down for two years by Magufuli under his recent Media Services Act.
This activity was then upstaged by the rapid revisions to the Tanzanian Mining Act dramatically increasing royalties, government ownership and prohibiting external arbitration and subsequently, by hitting Acacia with a US$190-billion tax and penalty bill.
However, the original source of controversy was the first committee report — considered technical in nature — that came up with the disputed numbers. It cited ranges and averages of concentrations of gold, copper, silver, sulphur, iron, iridium, rhodium, ytterbium, barium, tantalum and lithium that it measured in 44 of 277 impounded concentrate shipping containers, and concluded that they were understated or undeclared in Acacia’s concentrate assays.
The most striking were for gold, which they claimed ranged from 671 parts per million (ppm) to 2,375 ppm, and with an average 1,400 ppm gold that was 10 times higher than the 164 ppm that Acacia’s rigorous concentrate sale assays showed.
Acacia, its assayer SGS and the smelter company MRI Group defended the veracity of the original assays, but this large discrepancy raised a question in my mind of exactly how the committee had analyzed the concentrates, about which a rough translation from Swahili showed no indication.
The confidence or lack of qualification with which the numbers were stated suggested that they were not just a theoretical committee consensus, and instead must have come from an analytical source.
The unsupported high gold values reminded me of a situation in 2006 when I was running a drill program at Granduc in northwestern British Columbia. As part of our core-logging routine we scanned the core with a handheld Niton X-ray fluorescence (XRF) unit — the only manufacturer at that time after the Mars Rover missions — to produce a numerical log of copper and iron contents of the iron formation-hosted chalcopyrite-pyrrhotite mineralization.
The numbers supplemented our visual estimates and helped overcome backlogs at the geochem labs that were common in that era. The numbers looked good and were posted on the company website, and eventually corroborated by routine inductively coupled plasma (ICP) analyses.
The XRF unit was also extremely handy for elements like titanium and zircon, which would have required expensive whole rock analyses.
One day I found an arsenopyrite vein that I thought might be gold prospective, but was astounded when the XRF unit reported 2% gold. Repeated analyses with longer count times corroborated the high gold values, but my excitement ebbed when I realized that I could not see a trace of gold in the sample, when at 2% it should have been smeared all over the surface.
Concerned, I delved into the XRF analysis routine by downloading the raw data and energy dispersive spectral plot and found that the characteristic fluorescence gold peaks within the spectral range (the gold L alpha and beta) were buried under a mountainous arsenic K alpha peak at 10.5 keV (gold K-lines that might have been uncompromised are actually at 68.88 keV, far above the limits of the X-ray source).
The signals measured at the gold L-line energies were just part of the slope leading up to the major arsenic peak, and this was eventually confirmed by lab analysis of the rock, which returned nil gold.
The Acacia concentrates on the other hand have gold, however only at an average of 164 ppm, or 100 times less than the false peak I’d seen at Granduc (or perhaps more, if the committee measurements have any credence).
I wondered what the results might look like if someone used a handheld XRF to measure gold in the Acacia concentrate. Could peak overlaps from interfering elements account for the discrepancy between the committee and Acacia’s results, and do these interfering elements exist in the concentrate? Certainly, analyses of the concentrate by ALS show 700 ppm arsenic, 1,500 ppm zinc and 30 ppm tungsten, the main suspects in false gold peaks in XRF analyses, but can this account for the extra 1,200 ppm gold.
A decade after my Granduc experience, the handheld XRF industry has proliferated, with many new manufacturers aggressively competing to sell units to a range of industries, with claims for rapid identification and measurement of precious metals.
However, while there have been advances in higher resolution detectors, batteries and software leading to point-and-shoot capabilities, the essential physics remain the same (although a lot of the advertising suggests otherwise), and this creates limitations in many situations.
Most units use either a rhodium or silver target in their X-ray generating tube, which produces a continuous spectrum of X-rays ranging up to 50 keV. The X-ray counts are not equal for all energies and define a continuum curve (Bremsstrahlung) peaking at 5 keV and tapering to zero at 50 keV, and with sharp peaks at the characteristic emission lines for the source element, which for rhodium are at 22 keV.
The source X-rays hit the sample and through the delights of quantum mechanics cause fluorescence: the return of suites of X-rays precisely characteristic of each element present.
However, the counts of returned or fluoresced X-rays are variably proportional to source X-rays of similar and slightly higher energy, and the variable source spectrum calls for considerable corrections to convert counts of detected X-rays to measured concentrations.
On top of this, the sharp emission lines characteristic of the X-ray source element (e.g., the K-lines of rhodium) limits the useful range of analysis to below 20 keV.
This means that if you are measuring elements not much heavier than transition metals (but perhaps up to Z of 51 at Sb), the K fluorescence lines, which increase regularly with atomic number Z, are measured, while if you want heavier elements like gold, for which the K lines are out of range, then you have to rely on less distinctive L or M lines that are in the same energy range as K-lines of lighter and potentially more abundant elements.
This results in interferences that have to be resolved by peak stripping routines built into the units, and was the cause of the false gold report I observed in 2006.
Whether or not that situation would be avoided today by newer detectors and software, experienced exploration geologists do not expect to directly measure gold in complex geological materials, and instead use these units to measure associated pathfinder elements like arsenic.
Hence, it looked to me as though the spuriously high gold claimed by Magufuli’s committee might have been the result of using handheld XRFs, rather than standard fire assay techniques.
But were there any other indications? Iridium was the next big item on the committee’s list of grievously undeclared values in the concentrate, with US$49-million worth claimed to be hidden in the impounded containers.
A quick Internet search revealed several blogs about spurious iridium analyses by handheld XRFs, including one by Allan Fraser, a South African analytical chemist who had bought a mineral specimen online from Pasto Bueno in Peru, only to have it impounded by South African customs.
When he went to find out why, they confidently showed their handheld XRF measuring 3.45% iridium, 0.49% rhodium and 0.37% gold — plus some platinum and palladium — in his specimen.
The sample meanwhile was from a typical Andean vein system and consisted only of siderite, tetrahedrite, chalcopyrite and arsenopyrite, clearly not the right mineralogy or deposit type for iridium, which is never found in such high concentrations, even in the Bushveld complex.
Predictably, iridium falls in the same category as gold with its K-lines out of the energy range of handheld XRFs, resulting in reliance on L lines at 9.175 and 10.708 keV, which are interspersed with K-lines for both copper (copper K-beta at 8.91 keV) and arsenic (arsenic K-alpha 10.54 keV).
Measuring these overlapped minor peaks requires not only a high-resolution detector (SDDs can resolve peaks to 0.15 keV) but also a sound software library of relative peak intensities for all elements of interest, and an excellent routine for peak stripping.
This involves statistically adequate spectra for not just the major elements, but also the minor elements, and carefully stripping away all spectral features even remotely associated with the major constituents as well as predictable matrix effects, before attempting to attribute any concentrations to trace elements represented by the remaining minor peaks.
Short counting times (e.g., 15 seconds) may ephemerally show spurious elements that disappear from the handy on-screen report as the counting time increases, but from a human perspective, the initial report of an exciting element like iridium or gold raises expectations.
Could the iridium alleged to be 320 ppm in the concentrate be a false measurement by handheld XRF?
Both the theoretical peak overlaps (copper and arsenic) and the empirical case of the Pasto Bueno sample analysis by XRF suggest that it is highly likely, especially as copper has a significant concentration in the concentrate.
Rhodium (alleged to be 2 ppm) is a different case: if the analyzed material is reflective (metal or hard sulphides), X-rays characteristic of rhodium can bounce back to the detector at 20.22 and 22.71 keV. The detector cannot tell that these are just from the X-ray tube and chocks up rhodium as a measured constituent, perhaps only partly discounted by the software.
So, the presence of rhodium in the measurements, when none is expected, may be a subtle but direct indication that an X-ray device was used.
On top of this, rhodium L-alpha lines are down at 2.83 keV and may be overshadowed by the nearby K lines from sulphur (in pyrite and chalcopyrite), which is acknowledged at 36% in the Acacia concentrate.
Meanwhile, silver has K- and L- lines just above those of rhodium, as well as its L-lines in the neighbourhood of sulphur, which could readily augment the accepted silver content if not give a false silver peak — a known caveat.
Other elements reported by the committee that fall into the same handheld XRF measurement chaos as iridium include ytterbium and tantalum. Both have L-lines in the high-traffic spectral zone of copper and iron K-lines, which are known major elements in the Acacia concentrates at 14% weight for copper and 31% weight for iron.
Indeed, the only accurately reported elements by the committee seem to be for iron (27%) and sulphur (39%), representing the pyrite and chalcopyrite in the Bulyanhulu ore.
Oddly, the committee recommended these as a potential revenue source, like iridium, ignoring that smelters generally penalize for sulphur in concentrates and that the iron ends in slag.
Iron, sulphur and copper should all be reliably measurable by handheld XRF units, so it is a bit surprising that the copper contents were claimed to be understated. In fact, the committee report shows a wild range from 15.1% to 33.78% copper, sulphur from 16.7% to 50.8% and iron from 13.6% to 30.6%, suggesting either considerable variability in the Acacia concentrates or some rather imprecise analytical procedures on the part of the committee.
Their totals add up to 92%, which leaves insufficient room for the 10% silica from the quartz veins and 7.5% carbon from the sheared argillite units in which the gold-bearing quartz veins occur.
The overall implausibility of the concentrations claimed by the committee suggests they may result from many of the pitfalls and limitations inherent in the use of handheld XRFs, including matrix and operational effects (e.g., sum and escape peaks).
Nothing in the report indicates sampling protocols, analytical procedures, or even qualifications or identities of the first committee members.
However, the Tanzanian committee should not be immediately condemned for believing XRF results, if that is the case. Simplified operational modes encourage the same confidence for providing an accurate picture, as we get from point-and-shoot digital cameras.
Most of the emphasis of the report and its updated partner by the second committee focused on calculating the potential value of elements in the concentrates without questioning the veracity of the numbers used.
While I believe the committee acquired most of its implausible results from a handheld XRF, beryllium and lithium, which have low characteristic fluorescence X-ray energies, are undetectable by handheld units because their X-rays are attenuated in the air gap between the unit and the sample.
To get beryllium and lithium numbers the committee probably relied on some other rapid analytical technique available to them within the 20 days of their mandate, and selectively reported numbers that they judged representative.
Inductively coupled plasma atomic emission spectroscopy (ICP-AES) might be a reasonable choice but for the lack of any beryllium and lithium in the concentrate and the probability that iron, copper and sulphur would be over-limit by standard ICP analyses, while conveniently in range for a handheld XRF.
Perhaps a method particular to analyzing beryllium and lithium in gemstones — such as tanzanite and corundum, for which Tanzania is well known — were used, but that is a new mystery.
If indeed my speculation is correct — that the Tanzanian committee used handheld XRFs to analyze the concentrates — there should be some indication that XRF units were available to them.
Another Internet search revealed a document calling for bids by the Tanzanian government in early 2016 to supply handheld XRF units for use by government agencies.
More recently, articles in the Lusaka Times of neighbouring Zambia declared that they had implemented 18 “Olympus-X” portable XRF units for use in monitoring all mineral exports.
One Lusaka Times article stated: “Recently, there have been reports of mining companies exporting precious metals, especially gold, using falsified mineral content analysis reports. This has resulted in gross undervaluations of gold exports, thereby affecting revenue collection from mining companies.”
It is doubtful that Magufuli will elucidate the source of the disputed committee numbers, no matter how earnestly he and his committee believe in them. Certainly they will not stand up to scrutiny and would prove embarrassing regardless of the method used by the committee.
Likely, he used the report to raise patriotic fervour to back up his call for political solidarity and as a pretext to ram through objectionable amendments to the Mining Act in early July, while public expectations were high and Acacia and other miners were on the defensive.
Magufuli surely cannot believe that he can force payment of the claimed US$190 billion in back taxes and royalties, since revelation of faulty analysis by the government would collapse that argument quickly.
Instead, it is likely a bargaining tactic or ruse to force acquiescence with the more onerous and ambiguous conditions of the Mining Act amendments, and will be quickly discarded when Magufuli’s objectives are met.
While a president and national government may be within their interests and rights to push negotiations for more favourable resource revenue terms — and certainly Tanzania does not have good ones presently — the credibility and honour of the nation hinges on ensuring that its allegations are based on honourable procedures, and, in this case, credible analyses.
— Hardolph Wasteneys, P.Geo., APEGBC, is a consulting mineral exploration geologist with an engineering degree in mineral resource exploration (Queen’s, 1979) and PhD in mineral deposit research (Queen’s, 1990) based on study of epithermal silver mineralization in southern Peru. He did uranium-lead geochronology research at the Royal Ontario Museum during the 1990s on ore deposits and high-grade metamorphic terranes. He has worked in mineral exploration mainly in the Western Cordillera from the early 1980s to the present. He lives west of Campbell River, B.C., on Upper Campbell Lake. He can be reached at hardolph@gmail.com.
RELATED STORY:
Editorial: Barrick’s Acacia, Tanzanian gov’t stay on collision course
Great article from a technical perspective on the pitfalls of handheld XRF.
I remeber spotting a ‘Desert Dirt’ scam in the 1980s where desert varnish yielded ounce values of gold, silver, platinum and copper in material from a pit in Arizona. I calculated the sample size from the pit volume and concluded id was just the top few cm that were tested. The analytical method was Induction Coupled Plasma, a spectral technique. Primary gold platinum deposits are rare. You can get very high values of precious metals from copper and rust. Always use fire assays.